An excellent question about the accuracy of serology tests.

THREAD, WITH SPREADSHEET THAT YOU CAN DOWNLOAD AND PLAY WITH! https://twitter.com/delta_vee/status/1260928159521878016

I'm going to riff off this thread about a particular COVID-19 serology test with sensitivity 93.8%, and sensitivity 95.6%. https://twitter.com/taaltree/status/1248467732774785024

This means that if a person has had COVID-19, there is a 93.8% chance they test positive for antibodies. If the person has not had COVID-19, there is a 95.6% chance they test negative for antibodies. Pretty high numbers: this looks like a test that's around 95% accurate, right?

Alas, no, because the utility of the test depends not upon how likely it is that a person who has antibodies tests positive, but on how likely it is that a person who tests positive has antibodies, and similarly for negative results. And the numbers here are very different.

"But what are those numbers? ie - if I test positive for antibodies, how reliable is that result? In other words, how likely is it that I am actually immune to COVID-19?"

That's the key question. To answer it, we need to know the prevalence of COVID-19 in the general population.